Code and data you can pick up and use: lexicons, models and methods, all open. Below them, the replication code for individual studies. Papers themselves live on the Publications page.
Section 01
Word-level resources for measuring moral language at scale.
MoralStrength
Lexicon · Python package · Moral Foundations Theory
A moral foundations lexicon with crowd-annotated valence ratings — not just which foundation a word belongs to, but how strongly. It ships as a Python package that scores any text out of the box, so you go from raw text to moral foundation predictions in about three lines. It is the most reused thing we have built.
pip install moralstrength
LibertyMFD & LML — the Liberty lexicons
Lexicon · The moral foundation of Liberty
Moral Foundations Theory long had a hole in it. Liberty — the foundation behind anti-authority, anti-coercion and freedom-based moral reasoning — had no usable lexicon, which meant an entire dimension of political and vaccine discourse was invisible to automated analysis. These two lexicons fill it. They live alongside MoralStrength, under liberty/.
Section 02
Where a lexicon counts words, these read context.
MoralBERT
Fine-tuned language models · Ten moral dimensions
Transformers fine-tuned to detect moral values in social discourse. Context matters enormously for moral language, where the same word carries opposite valence depending on who is speaking. Ten models, one per moral dimension (Care/Harm, Fairness/Cheating, Loyalty/Betrayal, Authority/Subversion, Purity/Degradation), are on Hugging Face and load directly with transformers.
Tomea
Method · Explainable cross-domain comparison of moral rhetoric
A classifier can tell you that two communities talk about morality differently. Tomea tells you how. It compares what a moral classifier has learned across domains and surfaces the difference in human-readable terms — so you can interrogate a model instead of trusting it. Built with Enrico Liscio and the TU Delft group.
Section 03
Study-specific pipelines. Full citations are on the Publications page.
Section 04
If you are starting out in moral NLP, begin here rather than with the papers.
Capturing Human Values during Controversies
Tutorial · IC2S2 2023, Copenhagen
A hands-on tutorial for computational social scientists who want to measure moral and human values in text — slides, notebooks and worked examples, built around the lexicons and models above.
If something is broken, or a link gone stale just tell me —. If you are looking for code or data from a paper that is not listed here, ask: some of it can be shared on request.